一种基于支持向量机的模糊分类系统研究

提出一种基于支持向量机学习的模糊分类束纯模型．通过将支持向量机映射成等价的模糊分类系统,支持向量机的稀疏性表示等特性使得相应的模糊分类系统避免了“维数灾难”问题,并具有良好的泛化能力．另一方面,模糊系统的一些理论和应用成果也可用来进一步改善分类系统的性能．本文根据模糊集合的贴近度概念对模糊系统的语言变量进行约简,合并冗余的和不一致的模糊规则,然后采用粒子群优化方法改善模糊分类系统性能．该方法增强了系统的泛化能力,并可以理解为解决支持向量机中难以确定的系统参数问题的一种辅助方法．实验结果表明了该方法的可行性和有效性．

Fuzzy Classification System Based on Support Vector Machine

A fuzzy classification system model based on the Support Vector Machine is proposed in this research. Mapped from the Support Vector Machine equivalently, the model can avoid the "curse of dimensionality" and obtain the perfect generalization performance. On the other hand, fuzzy system theories and applied methods can also be used to improve its performance. Fuzzy rule optimization methods are developed to minimize the complexity of the model and improve the system generalization performance. The methods reduced the linguistic variables based on the similarity of fuzzy sets, mergered the redundant and inconsistented fuzzy rules. Finally, particle swarm optimization was used to adjust the system parameters effectively. The methods could be regarded as assistant measure to fixing the system parameters of the Support Vector Machine, which was one of the difficult problems in the Support Vector Machine. Experimental results showed that the methods are feasible and effective.